Solid State Drives And The Sort-Merge

Abstract

This paper discusses the history of the sort-merge routine and the impacts of hardware limitations on the performance of sort-merge processing.  The results of comparing a single-step sort-merge with a two-step sort-merge in a hard-disk drive (HDD) environment are presented to show that a two-step sort-merge can reduce total processing time. An evaluation is made of the total transfer time of three sort-merge variations without reference to seek time or rotational delay.  This evaluation prepares the statistics for application to the solid-state drive (SSD) environment, and the conclusion is that sort-merge routines that are optimized for the HDD environment are sub-optimal if applied to the SSD environment.  In addition, the sizes of the work files used by the three sort-merge routines are analyzed, and it is demonstrated that sort-merge routines that are optimized for the HDD environment will generate unnecessary wear if applied to the SDD environment.  Further, it is demonstrated that the key sorting routine should be preferred over the other sort-merge routines in a SSD environment.